Haiyan Super Typhoon: the brightest natural source of L-band radiation ever measured over the oceans
Typhoon Haiyan (known in the Philippines as Typhoon Yolanda) slammed into the Philippines in November 2013 with sustained winds of 310 kilometers per hour, making it one of the strongest tropical storms to date and the second-deadliest Philippine typhoon on record. Haiyan originated from an area of low pressure in the Federated States of Micronesia on November 2. Tracking generally westward, environmental conditions favored tropical cyclogenesis and the system developed into a tropical depression the following day. After becoming a tropical storm and attaining the name Haiyan at 0000 UTC on November 4, the system began a period of rapid intensification that brought it to typhoon intensity by 1800 UTC on November 5. By November 6, the Joint Typhoon Warning Center (JTWC) assessed the system as a Category 5-equivalent super typhoon on the Saffir-Simpson hurricane wind scale; the storm passed over the Palau shortly after attaining this strength.
Figure 1: Top: destructions after Super Typhoon Hayian passage over the Phillipines- Bottom: SMOS retrieved surface wind speed [km/h] along the eye track of super typhoon Haiyan from 4 to 9 Nov 2013.
SMOS intercepted the typhoon several times along its track. We selected only those passes were the signal was well detected and not too contaminated by RFI or land masses. As illustrated by Figure 1, this let one pass on the 4 as Haiyan was still a Tropical Storm, two on the 6th Nov (with the morning pass capturing only a small portion of the typhoon), one on the 7 prior landing towards Philippines and one interception on the 9, just before it passed over Vietnam. Passes on the evening of the 6 and during the 7th morning were close in time from the maximum intensity reached by that super storm (reached on the evening of the 7th).
As illustrated by Figure 2 top panel, the estimated excess brightness signal (First stokes parameter/2) due to surface roughness and foam-formation processes under the cyclone on the 7th morning overpass (i.e., after correcting for atmosphere, extra-terrestrial sources, salinity and temperature contributions) reached a record value of 41 K. To put such value in perspective of other natural oceanic signals, we plotted together the Tb jump measured during the passage of Hurricane Category 4-5 Igor in 2010, which was only 22 K! In contrast, global changes of surface salinity (32-38 pss) and temperature (0°C-30°C) only modify the Tb by ~5 K. So we believe such signal is very likely a natural extreme of sea surface emission at L-band over the oceans.
Figure 2: Top: North-south section trough the Haiyan Typhoon showing the change of residual brightness temperature (Th+Tv)/2 reconstructed from SMOS data at longitude of 130.05°E on the 7 Nov 2013 at 09:15Z Typhoon (black). The Blue curve is showing an equivalent section through the Igor Category 4 hurricane in 2010. The red line is illustrating the range of brightness temperature variation expected on earth due to sea surface salinity and temperature changes. Bottom: surface wind speed deduced from the excess brightness temperature.
Application of the wind-speed bi-linear retrieval algorithm that we derived in 2012 based on the established relationship between surface wind speed estimates during IGOR and the excess brightness temperature, we obtained the wind speed module shown in Figure 2, bottom. One can easily see that around the cyclone eye, wind speeds largely exceed the 64 knots threshold for typhoons within a more than 50 km radius. The spatial resolution of SMOS however does not allow to resolve the detailed wind speed structure around the eye. The maximum wind estimated from SMOS reach 142 knots !
Figure 3: Maximum sustained 1 minute wind speed estimated during Haiyan Typhoon. From SMOS data (black filled dots) compared to Advanced Dvorak Technique (ADT=blue diamond), CIMSS (yellow filled dots), SATCON (red) and Best Track from NHC (cyan). Note the empty circle correspond to the SMOS measurements for the 11/06 morning for which only a small portion of the cyclone signal was intercepted. Maximum 10 minutes wind speed deduced from SMOS algorithm were multiplied by 1/0.93, adopting the conversion factor proposed in (Harper et al., 2010) between one minute winds and 10 min winds.
Given the spatial resolution of SMOS, the wind speed measured is more equivalent to a 10 minute sustained wind than to a 1 minute one, traditionally used by forecasters in the US. Using a 0.93 conversion factor from 1 mn to 10 mn winds (Harper et al., 2010), 1 minutes sustained winds can be estimated from SMOS. The evolution of the maximum sustained wind speed deduced from SMOS is compared to other estimates in Figure 3. SMOS estimate compares well with standard methods. The Advanced Dvorak Technique (ADT) utilizes longwave-infrared, temperature measurements from geostationary satellites to estimate tropical cyclone (TC) intensity. This step-by-step technique relies upon the user to determine a primary cloud pattern and measure various TC cloud top parameters in order to derive an initial intensity estimate. It continues to be the standard method for estimating TC intensity where aircraft reconnaissance is not available (all tropical regions outside the North Atlantic and Caribbean Sea), however it has several important limitations and flaws. The primary issue centers upon the inherent subjectivity of the storm center selection and scene type determination proceedures. Secondly, learning the Dvorak Technique and its regional nuances and adjustments can take a significant time to master. SMOS winds in TC will be produced operationally next year so as to be ingested by numerical weather forecast model within 6 hours from acquisition. In regions not too polluted by RFI, this new impressive example shows again that SMOS new information shall well complement existing satellite observation to better forecast TC intensification and evolution.
Merged SMOS+AMSR-2 wind speed data
SMOS data provide a global coverage about every 3 days. During fast evolving storm events, SMOS swath can however miss interception with such fastly evolving storms or just capture a portion of the storm. In addition, SMOS data can be heavily contaminated in some areas by RFI, solar effects or land contamination. RFI are particularly problematic in the North west Pacific and in the Bay of Bengal. Combining SMOS and AMSR2 retrievals shall definitively help better characterizing high wind speed and storm events over the globe.
AS shown in Figure 3, for the Hayian case, SMOS estimate compares very well with standard methods. Nevertheless, the SMOS sampling along the complete life cycle of the storm is limited to 4 usefull overpasses. Complementing the SMOS sampling with other sensors would be therefore certainly beneficial.
With the recent developments of new methodologies to better retrieve surface wind speed in all weather conditions from X, C and L-band radiometer measurements from Space (Meissner and Wentz, 2009; El-Nimri et al., 2010, Reul et al., 2012, Zabolotskikh, 2013) the synergy of passive low-microwave frequency observations from space operating within the X to L-bands (AMSR2,WindSat, SMOS and SMAP) can now be envisaged. The complementarity and added-value with scatterometer ones (ASCAT & Oscat) and NWP products (ECMWF & NCEP) will be studied in the frame of our study with the aim to produce new blended surface wind speed products including the SMOS high wind speed data. As a first objective we plan to merge SMOS data and AMSR2 wind speed retrievals and probably further add the WindSat data and the future SMAP sensor ones. For AMSR2 high wind speed retrieval under rain, we will rely on a new methodology currently being developed by Zabolotskikh et al., 2013 . AMSR-2 is the Advanced Microwave Scanning Radiometer 2 on board GCOM-W1 satellite which substituted Aqua AMSR-E and was launched mid-may 2012.
An example of AMSR2 interception with Haiyan is shown in Figure 4 on 7 November 2013 at ~ 4:22 UTC.
Figure 4: Rain effects removal algorithm applied to AMSR2 X-band Tb for an overpass of super Typhoon Haiyan as the surface wind speed reached maximum values of 150 knts on the 7 Nov 2013.
SMOS intercepted Haiyan on the 7 Nov 2013 at 09:15Z while AMSR2 intercepted the Typhoon the same day about 5 hours sooner at ~ 4:22 Z. To compare the surface wind speed retrieved from both sensors, we recentered the eye estimated from each sensor data set based on the location of the maximum wind. Comparisons between both sensor surface wind retrievals are shown in Figure 5.
Figure 5 Top: Superimposed contours of SMOS (dashed) and AMSR2 (filled) surface wind speed fields estimated 5 hours apart as the sensors overpassed the super Typhoon Haiyan on the 7 Nov 2013. Middle: North-South and Bottom: East-West (right) sections of the retrieved wind speed through the storm (blue=SMOS; red=AMSR2).
SMOS is operated at 1.4 GHz while data used from AMSR2 involve 7 and 11 GHz channel data and the respective algorithms used to retrived surface wind speed are very different in nature (mono-frequency and no rain correction for SMOS, rain and multi-frequency data fro AMSR2). Nevertheless, the comparisons shown in Figure 5 reveal that above hurricane force (>33 m/s) both instrument see very similar wind speed structures. Major differences are observed in the lowest wind speed range below hurricane force. It can be due to temporal evolution of the wind field in between the two observations or to differences in the breaking wave, sea state, spray or other geophysical impact on the brightness temperatures. Nevertheless, the consistency between both sensors in the high wind speed regime is very impressive and promising for the generation of new low-frequency microwave radiometer merged high wind products.
Figure 6: a) Contours of surface wind speed at 34, 50 and 64 knots retrieved during the passage of super Typhoon Haiyan in Nov 2013 a) from SMOS sensor, b) from AMSR-2 and c) by merging SMOS and AMSR-2 data.
Figure 6 is further illustrating the strength of the synergies and data merging between these two sensors in term of increased spatial and temporal coverage for rapidly evolving and intense storms such as Haiyan typhoon.
Figure 7:Contours of the merged SMOS+AMSR2 retrieved winds over Haiyan at the threshold levels of 34 (blue), 50 (green) and 64 (orange) knots.
Figure 8: Maximum sustained 1 minute wind speed estimated during Haiyan Typhoon. From SMOS data (black filled dots) and AMSR2 (black filled squares) compared to other top-of the atmosphere measurements. Note the empty circlesand squares correspond to the SMOS or AMSR2 measurements for which only a small portion of the cyclone signal was intercepted.
As shown in Figures 6-8, by combining both sensors, consistent and more continuous estimations of key parameters for describing the storms in the context of improving NWP forecasts, such as radii at 34, 50 and 64 knots, and maximum sustained winds can now be provided and augmented.